Validating Econometric Design: A Case Study from James Habyarimana's Updates to his 2003 Job Market Paper
causalinf.substack.com
This substack takes a closer look at a significant yet subtle transition in economic research methodology using James Habyarimana’s 2005 revision of his 2003 job market paper as a case study. James, who graduated from Harvard and is now a professor at Georgetown, initially presented a difference-in-differences (diff-in-diff) analysis (including an event study) of the closure of four Ugandan banks in 1997 and 1998 on firms’ performances who were linked to those banks. It is an interesting study on its own. But for this substack, I am using it as a case study because his 2003 paper did not use the phrase “parallel trends” but did in the 2005 revision. As we all know, “parallel trends” is the key identifying assumption in diff-in-diff, so the fact that it was absent in the original 2003 but is present in the 2005 revision is interesting for helping understand the broader paradigm shifts that causal inference brought to the profession.
Validating Econometric Design: A Case Study from James Habyarimana's Updates to his 2003 Job Market Paper
Validating Econometric Design: A Case Study…
Validating Econometric Design: A Case Study from James Habyarimana's Updates to his 2003 Job Market Paper
This substack takes a closer look at a significant yet subtle transition in economic research methodology using James Habyarimana’s 2005 revision of his 2003 job market paper as a case study. James, who graduated from Harvard and is now a professor at Georgetown, initially presented a difference-in-differences (diff-in-diff) analysis (including an event study) of the closure of four Ugandan banks in 1997 and 1998 on firms’ performances who were linked to those banks. It is an interesting study on its own. But for this substack, I am using it as a case study because his 2003 paper did not use the phrase “parallel trends” but did in the 2005 revision. As we all know, “parallel trends” is the key identifying assumption in diff-in-diff, so the fact that it was absent in the original 2003 but is present in the 2005 revision is interesting for helping understand the broader paradigm shifts that causal inference brought to the profession.